Editorial principles

The rules we follow when writing and reviewing pages.

This page is the longer version of the editorial note on our About page. It is meant for readers, contributors, and other reference sites who want to understand our standards.

Transliteration

Arabic does not transliterate cleanly into the Latin alphabet, and there is no single transliteration scheme that everyone uses. The academic schemes (IJMES, Library of Congress, EI3) are precise but unfamiliar to most readers; the popular schemes are inconsistent. We use a single readable scheme across the site, with these conventions:

• Long vowels are written doubled: aa, ii, uu. We do not mix ee/ii or oo/uu.
• Emphatic consonants are capitalized: S (ص), D (ض), T (ط), Z (ظ).
• H is the emphatic ح; lowercase h is the lighter ه.
• ع (ʿayn) is written ʿ. ء (hamza) is written ', and is omitted at the start of a word.
• خ is kh, غ is gh, ق is q, ش is sh, ث is th, ذ is dh.

Arabic always appears alongside the transliteration. The transliteration is meant as a guide for the early learner; it is not a substitute for reading the script, and we never use it without the script.

Dialect labels

Every phrase, idiom, or vocabulary item is labelled by register: MSA (Modern Standard Arabic), Egyptian, Levantine, Gulf, Maghrebi, Iraqi, Sudanese, or Yemeni. Where a phrase is broadly understood across the Arab world but only spoken naturally in some places, we say so. We resist the urge to give a single "Arabic" version when the underlying reality is regional.

Sourcing

For etymological claims (especially in Loanwords) we cite a word as derived from Arabic only when standard etymological dictionaries — the OED, Corriente, the Diccionario de la Real Academia Española — agree on the path. For historical and cultural claims, we prefer authority that names its sources over authority that doesn't. Where the path is contested, we say so.

Human review

AI tools are used in the drafting and checking of pages, in the same way most editors and writers in 2026 use them. No page is published without a human editor reading it through and signing off. Pages flagged with corrections are re-edited by a human, not by a model alone. If a page contains an error, the responsibility is the editors'.

What this site won't do

• No AI-generated final copy without human review. Every page has a human editor of record.
• No affiliate links. If a book or app is recommended, no one is paid for it.
• No advertising beyond what is declared in our ads.txt.
• No tracking. No analytics scripts, no pixel-counting widgets, no third-party JavaScript loaded for surveillance. The reference pages are plain HTML and one stylesheet.
• No chat or "ask the AI" boxes. The site is a reference, not a conversational interface.

The translator on the homepage

The homepage carries a small translator widget that sends the typed text to the free MyMemory translation API and shows the result. It is there because readers often arrive looking for a quick translation; sending them somewhere else would be unhelpful. The widget is a single piece of inline JavaScript on the homepage only — every other page on the site is plain HTML. The translator is a convenience, not a substitute for the reference: machine translation flattens register and dialect, which is precisely what the rest of this site is built to preserve.

Corrections and updates

Pages carry the date they were last edited in the footer. Substantive corrections — anything beyond a typo — are noted in a short revision line at the bottom of the affected page when the change is made. Routine copy-edits and clarifications are not noted individually.